ITC 2019: International Timetabling Competition

NEW: Journal Paper

Our journal paper about the International Timetabling Competition 2019 is finally published:

Tomáš Müller, Hana Rudová and Zuzana Müllerová, Real-world university course timetabling at the International Timetabling Competition 2019. Journal of Scheduling, DOI 10.1007/s10951-023-00801-w, 2024. [Paper] [PDF] [Citation]

ITC 2019 Overview

Building on the success of the earlier timetabling competitions, the International Timetabling Competition 2019 is aimed to motivate further research on complex university course timetabling problems coming from practice.

Our goal is the creation of rich real-world data sets. Thanks to the UniTime timetabling system, we collect a strong set of data with diverse characteristics. The key novelty lies in the combination of student sectioning together with standard time and room assignment of events in courses. To make the problems more attractive, some of the less important aspects of the real-life data are removed while retaining the computational complexity of the problems.

Three groups of data instances will be published during the competition (early, middle, late). While the order of the competitors will be based on all the published data instances, the instances released later in the competition have a much higher weight in the final ranking. The ranking is based on the computation of points in the F1 championship. Additional test data instances are available from the announcement of the competition in August 2018.

The competition is supported by the PATAT conference and by the EURO working group on Automated Timetabling (EWG PATAT). The winner will get a cash price of 1000 EUR, the second place will be awarded 500 EUR, and the third place 250 EUR. In addition, we will provide a free registration to the PATAT 2022 conference for one member of each team of the first, the second and the third place submissions. In order to receive the prizes, the three best competitors are required to attend the PATAT 2022 conference and present their successful results in the Special Track that will be dedicated to the competition. In addition, they will be expected to submit a full version of their papers to the special journal issue of the conference. Additional prizes will also be provided during the run of the competition. The first, the second and the third best competitor on February 1 and June 1 will be announced and obtain 300 EUR, 200 EUR, and 100 EUR, respectively. We will publish the quality of the best computed solutions from February 1 and June 1. Thanks to our sponsor, the company ORTEC, the best solutions for the late instances will be awarded by 150 EUR per instance. When there are multiple best solutions of the same quality submitted, only the first submitted solution will be rewarded. Thanks to the Apereo Foundation, the best open source solver will be awarded a price of 500 USD. Such a solver will need to be available in a public repository under an open source license with information on how to compile and run the solver on the competition instances.

A web service will validate the solutions for the competition problems and will allow for the valid solutions to be uploaded to the website. The best-uploaded solution for each competitor and instance will be used for the ranking of the competitors, and it will be published on the competition website once the competition is concluded. The website will be maintained even after the competition to keep track of new solutions, publications about them, and include new data instances.

While the competition is based on the ideas of the previous competitions, some important aspects are different. In particular, the time needed to compute solutions is not constrained, and neither is the number of CPU cores or machines that the solver can use. At the same time, we do not expect to run solvers of the competitors on our hardware, i.e., solvers built on commercial software are allowed. We may ask to see the source code of solvers for the competitors.

The Organization section contains more information about the organization of the competition and the Rules section explains the competition rules and submission details. The Format section gives the description of the XML data format used in data instances. Detail description of the competition including precise specification data formats is available in the following paper:

Tomáš Müller, Hana Rudová and Zuzana Müllerová, University course timetabling and International Timetabling Competition 2019. In PATAT 2018 - Proceedings of the 12th International Conference on the Practice and Theory of Automated Timetabling (PATAT 2018), pages 5-31, 2018. [Paper] [Presentation] [BibTeX]

Problem Overview

The competition problem consists of university courses and student course requests. A course may have one or more classes such as one lecture and several seminars. A hierarchical course structure is defined to specify how students attend classes. One class can have several meetings a week and must be scheduled at the same time and in the same room. It can be taught at either all or selected weeks of the semester. For example, a course may have two seminars of up to 20 students, and each seminar can take place on Mondays and Wednesdays at 9 am every other week, one seminar on odd weeks, the other on even weeks.

The aim is to find a proper time, a room and students for all the classes. Each class has a set of acceptable time and room assignments, together with non-negative penalties that define their suitability. All students must be correctly assigned to classes of their requested courses. Student conflicts, which occur when a student cannot attend a pair of classes (e.g., they overlap in time), are to be minimized. Distribution constraints express relations among a set of classes, e.g., seminars must be taught after the lecture each week. Distribution constraints may be hard or soft. Hard constraints must be satisfied, soft constraints are associated with a penalty.

A feasible solution exists for each data instance and only feasible solutions are accepted in the competition. The overall solution cost should be minimized. It is computed as a weighted sum of all the time and room assignment penalties, penalties of broken soft distribution constraints, and the total number of student conflicts.

Interim Results

We are pleased to announce the winners of the first and the second milestones.

Competition Winners

We are pleased to announce the following winners.

Dennis S. Holm, Rasmus Ørnstrup Mikkelsen, Matias Sørensen, Thomas R. Stidsen
MaCom / Technical University of Denmark, Denmark
Best solutions of nine late instances (all except agh-fal17).

Efstratios Rappos, Eric Thiémard, Stephan Robert, Jean-François Hêche
HEIG-VD, Switzerland

Edon Gashi, Kadri Sylejmani
University of Prishtina, Kosovo
Winner of the best open-source award.

Karim Er-rhaimini
Ministère de l'éducation nationale, France
Best solution of one late instance (agh-fal17).

Alexandre Lemos, Pedro T Monteiro, Inês Lynce
INESC-ID / IST, Universidade de Lisboa, Portugal

Congratulations! More details are available on the Score page. All the uploaded instances can be downloaded from the Results page.

The award ceremony was held online on September 2, 2020, using Zoom [Presentation] [Video].